Friday, March 14, 2014

Analyzing publications

I have an interest in analyzing publications.  In prior work, we looked at clues authors leave in the latex source of papers (not PDFs).  We extended that to study if there is a bias how one or the other authors' names is used to refer to a joint publication. Beyond that, we got curious about a natural challenge. We write reference letters for other researchers and while recommending Alice, we need to find Bob and Carol who are suitable, comparable researchers. This is not a  trivial task. Sometimes one looks across areas and other times within; sometimes one looks for someone with similar experience and other times with a similar initial trajectory so we can project out by analogy; sometimes one is comparable in volume, other times by citations and quality; etc. We did a simple study recently to see if one can automatically/algorithmically find comparable researchers. Here is the paper, we found some examples that challenged us to explain them, please email me if you want to find out more.

Anyway, this is the backdrop to what I really wanted to blog (v):

Once in a while we get rankings of universities (ex). These cause angst or jubilation.  On the other hand, there is a lot of theory of rankings, axiomatic, algorithmic, strategic and so on that point to the challenge of producing defensible rankings.  Going beyond, are there interesting research ideas to automatically/algorithmically produce university rankings? I would love to see some research that even tackles a narrower problem of ranking departments based on the publications of the current faculty, or ranking based on the graduated PhDs, etc. 



Blogger Suresh Venkatasubramanian said...

I think a major problem with automating the process of rankings is not the algorithms so much, but the problem of collecting reliable data. The NRC debacle showed how bad things can get.

5:18 PM  

Post a Comment

<< Home